This is the mail archive of the
glibc-bugs@sourceware.org
mailing list for the glibc project.
[Bug math/20024] vectorized sincos trashes the stack
- From: "andrew.n.senkevich at gmail dot com" <sourceware-bugzilla at sourceware dot org>
- To: glibc-bugs at sourceware dot org
- Date: Wed, 11 May 2016 16:30:47 +0000
- Subject: [Bug math/20024] vectorized sincos trashes the stack
- Auto-submitted: auto-generated
- References: <bug-20024-131 at http dot sourceware dot org/bugzilla/>
https://sourceware.org/bugzilla/show_bug.cgi?id=20024
--- Comment #5 from Andrew Senkevich <andrew.n.senkevich at gmail dot com> ---
Jakub, but fixed so implementations can be slow enough.
Linear clause for attribute __simd__ can be added for GCC 6.2 and used under
that version check in Glibc headers.
Joseph, can we fix ABI with according change in declaration for sincos?
Both ways are subject for backport to 2.22 and 2.23 releases, but I believe
amount of existing binaries having wrong vector sincos calls is zero (else it
was reported earlier).
Also question what about to have 2 different simd declarations (w/ and w/o
linear)?
Currently GCC doesn't vectorize sincos in a common usage case. Only quite a
tricky testcase involving vector_size attribute may result in vectorized sincos
that invalidates testing unfortunately.
--
You are receiving this mail because:
You are on the CC list for the bug.